ACF is a method for object detection which approximately constructing a dense feature pyramid used for Adaboost classifier. Our work focuses on this method and implement the whole detection process on heterogeneous hardware platform. This design achieves a detection performance of 134 fps consuming less hardware resources.